metadata
license: apache-2.0
Hello darkness, my old friend I've come to talk with you again.
IMPORTANT: Make sure you have the latest version of llama.cpp to use these:
- The PR that allows multiple control vectors to be loaded (without zero-padding) just got merged today (27/06/24).
- Older versions of
llama.cpp
will just silently load the first control vector (and none of the others) if the layer index of the final direction in each file does not match...
To use these control vectors effectively you will need to use the "--control-vector-scaled
" option like this:
llama-cli --model <model name>.gguf --control-vector-scaled <model name>-dark.gguf 0.5 --control-vector-scaled <model name>-chaos.gguf 0.5 [the rest of your CLI arguments...]
or:
llama-cli --model <model name>.gguf --control-vector-scaled <model name>-dark.gguf 1.0 --control-vector-scaled <model name>-chaos.gguf 0.0 [the rest of your CLI arguments...]
or:
llama-cli --model <model name>.gguf --control-vector-scaled <model name>-dark.gguf 0.5 --control-vector-scaled <model name>-chaos.gguf 0.25 [the rest of your CLI arguments...]
NOTE:
- Use positive scale factors to make the model "more dark" or "more chaotic".
- I suggest you use
--control-vector-scaled 0.5
and--control-vector-scaled 0.5
to start and then test the effect. - The "chaos" control vectors generally seems less effective than the "dark" control vectors.
- Some models like
command-r:35b
andcommand-r-plus:104b
need lower scale factors, whereasmiqu-1:70b
seems to need (much) higher scale factors to stamp out pesky redemption arcs. - You can use one control vector file alone if you want, or alternatively set the scale factor to
0.0
for traits you don't want to use. - You can use the same "
--control-vector-scaled
" command line arguments for "llama-server
" as in the above "llama-cli
" examples.